Authors: Matthew Williams
Mentors: Elena Laricheva
Insitution: Utah Valley University
Protein-ligand docking is a computational method widely used in drug discovery to predict binding affinities of small molecules to target receptors. However, despite its widespread utility, the method has inherent limitations that can lead to false negative and false positive results, affecting its reliability. False positives occur when docking predicts strong binding affinities that experimental evidence does not confirm, while false negatives arise when the method fails to identify potent binders validated in experiments.
Various factors contribute to these inaccuracies, including limitations in scoring functions and search algorithms, but a significant issue lies in the neglect of protein dynamics, i.e., receptor flexibility. To address this limitation, flexible docking methods, which partially account for receptor flexibility, have been developed, but they come at a considerable computational cost. In this project, we incorporated molecular dynamics simulations and k-means clustering to improve prediction of binding energies of a series of small molecules to the human dopamine 2 receptor, a crucial therapeutic target for neuropsychiatric disorders.
Our findings demonstrate that sampling conformational states through molecular dynamics and clustering, followed by docking to representative clusters, offers a more accurate assessment of binding energies. Remarkably, this enhanced predictive capability is achieved with minimal additional computational expense.